Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Claim. Przemek Chojecki's post of 16 April 2026 in the discussion thread of Problem 326 links a short note with a counterexample that the post credits to GPT-5.4. The erdosproblems.com wording then asked Erdős and Graham's "another way of stating the problem": does every basis of order 22 have a subset which is also a basis and for which lim⁡ak/k2\lim a_k/k^2 does not exist? The note answers no. Its basis is the set of natural numbers whose ternary digits are all 00 or 11, a basis of order 22 since the sum of two such numbers can have any ternary digit, and every sub-basis {bk}\{b_k\} of it has bk/k2→0b_k/k^2\to0, so the limit exists for every sub-basis.

Submission note. Posted to the erdosproblems.com forum by Przemek Chojecki on 16 April 2026:

Looks like this is false. A counterexample given by GPT-5.4 is

>A={∑j≥0εj3j:εj∈{0,1},> for all but finitely many j}>> A = \Bigl\{ \sum_{j\ge 0} \varepsilon_j 3^j : \varepsilon_j\in\{0,1\}, > \ \text{for all but finitely many j} \Bigr\} >

Thus AA is the set of natural numbers whose ternary digits are all 0 or 1 (alternatively exclude 0 depending on the convention). Here's a short note with a proof that this is indeed a counterexample

Why it is rejected. It answers a variant, not the Statement. The note answers erdosproblems.com's earlier wording, in which a limit 00 of ak/k2a_k/k^2 counts as a limit. The erdosproblems.com curator replied the same day that the limit was meant to be a constant c≠0c\ne0, since otherwise there are easy counterexamples such as this one, and rewrote the problem as Erdős and Graham's minimal-basis question, the Statement of Problem 326: is there a minimal basis of order 22 with ak/k2→ca_k/k^2\to c for some c≠0c\ne0? A basis whose sub-bases all have bk/k2→0b_k/k^2\to0 has no such sub-basis and says nothing about minimal bases with a nonzero limit, so its mathematics answers the variant and does not settle this problem; the value disproved is its answer to that variant.

Standing. Rejected, as answering a variant. The note is unrefereed, the thread records no review of it, and this corpus has not checked it.

Depends on. No page of this wiki.